Constructor is always looking for new ways to enhance the shopping experience for our customers’ shoppers. Our latest facet group sorting algorithm update introduces machine learning to predict the most relevant order for facet groups based on multiple dynamic features.

What is it?

A new machine learning-based approach to facet group sorting that incorporates more data-driven factors, such as coverage and facet type. Instead of using a hard-coded formula, the algorithm uses a trained machine learning model to predict the optimal order of facets for each search or browse query in real-time.

How does it work?

The machine learning model considers the following features:

• Popularity: How often each facet is used by shoppers over a recent period.
• Relevance Attractiveness: The appeal of top items within each facet.
• Diversity: The distribution of items within each facet.
• Coverage: The percentage of products that have the facet.
• Facet Type: The type of facet being displayed.
  
  

These features are processed in real-time by a model trained on historical data. Using this data, the model predicts a final score for each facet, which is then used to dynamically rank the facets during search and browse queries

What problem does it solve?

The new model-based approach enhances the responsiveness and accuracy of facet sorting. It allows for real-time adjustments based on current usage and shopper preferences, ensuring that shoppers always see the most relevant and attractive facets first, leading to a better shopping experience.

Availability:

The feature is rolling out to all customers, replacing the old method of facet relevance sorting with this machine learning-powered approach.

For more information, please contact your Customer Success Manager or reach out to us at support@constructor.io